Search

Semiconductor manufacturing has been extensively exploiting machine-learning (ML) to process equipment sensory data (ESD) for near-real time anomaly detection (AD). ESD characteristics are highly diversified and data lengths vary among processing steps and cycles. Cleansing ESD with minimum distortion (CMD) to fit the fixed-length input requirement by ML-based AD is critical to AD effectiveness and is challenging. This paper presents a novel CMD method of four innovations: i) statistical mode-based equalization of step data lengths for the least number of step data length changes, ii) importance indicator value (IIV) of a data sample based on its relative difference with the subsequent sample, and iii) step data segmentation into groups based on samples of significant IIVs and the least-entropy-group-to-cleanse-first rule, and iv) cleansing the least IIV sample(s) in the selected group for step data length equalization. CMD application to ESD demonstrates its characteristics preservation property. Simulation experiments are on an integration of data cleansing with an unsupervised ML-based AD system, STALAD. Comparisons with two benchmark methods over AD scenarios of small-scale drifts and shifts show that CMD not only is superior in facilitating accurate detection by STALAD but also helps detect anomaly much earlier than using the two benchmarks.

Background Polyethylene glycol (PEG) is widely used in industry and medicine. Anti-PEG antibodies have been developed for characterizing PEGylated drugs and other applications. However, the underlying mechanism for specific PEG binding has not been elucidated. Methods The Fab of two cognate anti-PEG antibodies 3.3 and 2B5 were each crystallized in complex with PEG, and their structures were determined by X-ray diffraction. The PEG-Fab interactions in these two crystals were analyzed and compared with those in a PEG-containing crystal of an unrelated anti-hemagglutinin 32D6-Fab. The PEG-binding stoichiometry was examined by using analytical ultracentrifuge (AUC). Results A common PEG-binding mode to 3.3 and 2B5 is seen with an S-shaped core PEG fragment bound to two dyad-related Fab molecules. A nearby satellite binding site may accommodate parts of a longer PEG molecule. The core PEG fragment mainly interacts with the heavy-chain residues D31, W33, L102, Y103 and Y104, making extensive contacts with the aromatic side chains. At the center of each half-circle of the S-shaped PEG, a water molecule makes alternating hydrogen bonds to the ether oxygen atoms, in a similar configuration to that of a crown ether-bound lysine. Each satellite fragment is clamped between two arginine residues, R52 from the heavy chain and R29 from the light chain, and also interacts with several aromatic side chains. In contrast, the non-specifically bound PEG fragments in the 32D6-Fab crystal are located in the elbow region or at lattice contacts. The AUC data suggest that 3.3-Fab exists as a monomer in PEG-free solution but forms a dimer in the presence of PEG-550-MME, which is about the size of the S-shaped core PEG fragment. Conclusions The differing amino acids in 3.3 and 2B5 are not involved in PEG binding but engaged in dimer formation. In particular, the light-chain residue K53 of 2B5-Fab makes significant contacts with the other Fab in a dimer, whereas the corresponding N53 of 3.3-Fab does not. This difference in the protein-protein interaction between two Fab molecules in a dimer may explain the temperature dependence of 2B5 in PEG binding, as well as its inhibition by crown ether.

Background: Development of an efficacious egg-free mock-up H5N1 vaccine is key to our preparedness against pandemic avian flu. Methods: This is a single-center, randomized, observer-blinded phase I clinical trial evaluating the safety and immunogenicity of an alum-adjuvanted Madin–Darby canine kidney (MDCK)-derived inactivated whole-virion H5N1 influenza vaccine in healthy adults. Hemagglutination inhibition (HAI) and neutralizing antibody titers were measured using horse and turkey red blood cells (RBCs). Results: Thirty-six adult subjects were randomized to receive two doses of 0.5 mL of the MDCK-derived H5N1 alum-adjuvanted vaccine containing 7.5, 15, or 30 μg of hemagglutinin (HA) 21 days apart. The candidate vaccine was well tolerated and safe across the three dosing groups. The most frequent adverse event was injection site pain (46.5%). Both HAI and neutralizing antibody titers increased after each vaccination in all three dosing groups. The best HAI responses, namely a seroconversion rate of 91.7% and a geometric mean ratio of 9.51 were achieved with the HA dose of 30 μg assayed using horse RBCs at day 42. HAI titers against H5N1 avian influenza virus was significantly higher when measured using horse RBCs compared with turkey RBCs. Conclusions: This Phase I trial showed the MDCK-derived H5N1 candidate vaccine is safe and immunogenic. The source of RBCs has a significant impact on the measurement of HAI titers (ClinicalTrials.gov number: NCT01675284.). Keywords: Influenza A, Avian influenza, H5N1, Vaccination, Immune response